AI image tools DreamUp, Midjourney and DreamStudio are at the center of a proposed class action that alleges the products’ reliance on copyrighted images to generate work “in the style of” certain creators has illegally siphoned money from countless artists.
The 46-page lawsuit was filed in California on January 13 by three full-time artists who allege defendants Stability AI LTD., Midjourney and DeviantArt are guilty of “blatant and enormous” copyright infringement as the companies have pulled in significant profits from artificial intelligence products that are “powered entirely” by the hard work of artists nationwide.
Want to stay in the loop on class actions that matter to you? Sign up for ClassAction.org’s free weekly newsletter here.
The lawsuit scathes that neither Stability nor DeviantArt has attempted to negotiate licenses for any of the copyrighted images that were allegedly scraped from the web and used without permission to create, or “train,” the artificial intelligence their respective tools use. At the same time, neither company shares any of the hefty revenues they enjoy from AI-generated art with the artists from whom the reproduced work is derived, the suit says.
Similarly, Midjourney, whose product, like DreamUp and DreamStudio, allows anyone to use text prompts to produce a digital image via AI, relies on the wrongful appropriation of “millions of copyrighted images created by artists,” the complaint says.
According to the case, the harm experienced by artists is “not hypothetical” as AI-generated images are increasingly sold online—and have even been submitted as winning entries in art competitions. This week, stock photo juggernaut Getty Images took on Stability AI when it filed a lawsuit alleging that the company illegally scraped and processed millions of images from its repository website to train its software, violating copyrights in the process.
The lawsuit alleges Stability AI, Midjourney and DeviantArt have run afoul of the Digital Millennium Copyright Act (DMCA) and California consumer protection laws.
“In the style of ...”
To create its widely used Stable Diffusion software, the backbone of all three products at issue in this suit, London-based Stability downloaded or otherwise acquired copies of “billions” of copyrighted images without permission, the filing states.
According to the case, these images – which the suit rightfully dubs as “training images” – were then essentially used to teach Stable Diffusion to produce seemingly new pictures without the consent of or compensation to the original creating artists, the case relays.
These purportedly “new” images—allegedly developed from the “over five billion” taken from websites beyond count—are based entirely on the unlicensed “training” images and are derivative of the original works from which Stable Diffusion draws when assembling an AI-generated image, the suit contends.
The lawsuit summarizes that the diffusion technique used by the software ultimately allows for a machine-learning model to calculate how to remake a copy of a training image. Notably, a reconstructed copy of an image will not perfectly match the original as some “small, unimportant, or insignificant details” may be lost as the data is crunched into a smaller size, the case points out.
“The resulting image is necessarily a derivative work, because it is generated exclusively from a combination of the conditioning data and the latent images, all of which are copies of copyrighted images,” the lawsuit contends, alleging Stability “simply took” from creators the all-important training images, many of which came from the DeviantArt online community, from which it now derives massive profits. “It is, in short, a 21st-century collage tool.”
Since Stability released Stable Diffusion under an open-source license last summer, the software has been rapidly adopted in the increasingly lucrative AI image-generating space, the complaint shares.
Critically, the AI images pumped out by Stable Diffusion, whether through Stability, Midjourney or DeviantArt’s tools, go head to head in the marketplace with artists’ original images, the complaint explains. Previously, when a buyer sought a new image “in the style of” a particular creator, they would have to pay a commission or license an original image from that artist, the suit relays. Now, however, purchasers can essentially input an artist’s name to generate a new, close-enough-to-the-original image, completely side-stepping the crucial task of paying the artist for their work, the lawsuit alleges.
As the suit tells it, only a select few artists are skilled enough to reproduce art that is convincingly in another artist’s style, let alone do so for vast numbers of artists.
“AI Image Products do so with ease by violating the rights of millions of artists,” the case summarizes.
The filing goes on to allege that DeviantArt, specifically, has “betrayed its artist community” by declining to attempt to negotiate licenses for any of the training images it uses to generate AI art. Although the company purports to let users create AI art with the knowledge that “creators and their work are treated fairly,” this is “false and misleading,” the lawsuit alleges.
“Rather than standing up for the rights of its members by rejecting Stable Diffusion and other sources of AI-generated art, DeviantArt has gone the opposite direction: it has built an app called DreamUp that is based on Stable Diffusion,” the suit reads, calling DeviantArt a “co-conspirator” in the illegal use of copyrighted materials. Further, the case alleges, DeviantArt could have taken legal action against Stable Diffusion for violating its terms of service (e.g., by reproducing, distributing or publicly displaying derivative works based on the work of copyright holders), yet chose not to do so.
Additionally, despite acknowledging in its terms of service that data scraping is considered an “unauthorized use” of artists’ work, the company claims that original artists are responsible for “policing their own works,” the suit says. As the complaint tells it, this amounts to DeviantArt “washing its hands of the matter,” a stark about-face to its “professed interest” in using its terms of service to shield the artist community.
“Instead of standing up for artists and using its resources to combat illegal AI data scraping, [DeviantArt] is forcing artists to take matters into their own hands,” the lawsuit reads.
Who’s covered by the lawsuit?
The suit looks to represent all persons or entities nationalized and/or living in the United States who own a copyright interest in any work that was used to train any version of an AI image product that was offered directly and/or incorporated into another product by Stability AI LTD., Midjourney and/or DeviantArt within the relevant statute of limitations period.
The case also looks to cover those who maintained an account on DeviantArt, posted copyrighted work on DeviantArt and had that work used to train any version of an AI image product.
I’m an artist whose copyrighted work was used without permission for AI-generated images. How do I get involved?
When a proposed class action is first filed, there’s usually nothing a person needs to do to join or sign up for the case. The time for a “proposed class member”—i.e., someone who’s covered by a lawsuit—to take action is typically if and when a case settles, which could take months or years.
If you’re an artist or creator whose copyrighted work has been misappropriated to produce AI-generated images, or simply want to stay in the loop on class action lawsuit and settlement news, sign up for ClassAction.org’s free weekly newsletter.